Embed Size (px)
DESCRIPTION
Intro slides for the EventLogging Workshop, introducing a new infrastructure built by the Wikimedia Foundation for web analytics and collaborative data modeling. http://www.mediawiki.org/wiki/EventLogging/Workshop
Citation preview
EventLogging WorkshopDario Taraborelli Ori Livneh Maryana Pinchuk
[email protected] [email protected] [email protected]
Wikimedia FoundationMarch 7, 2013
http://bit.ly/ELWorkshop
Program1.30-3.00. EventLogging tutorial
• Intro and rationale (Dario)• From an idea to a data model: Designing a schema (Maryana)• Instrumenting code, collecting and validating EventLogging data (Ori)• Accessing, QA'ing and analyzing EventLogging data (Dario)
3.00-3.30 Coffee break
• Coffee, cookies and snacks
3.30-5.00 Hacking session
• Hands-on session to learn how to use EventLogging
Service announcements
IRC channel #wikimedia-e3 @ freenode.net
Streaming/recordingYou're on the air (1.30-3pm)http://www.youtube.com/watch?v=ZBuTjl0lh9Y
Mailing listSign up at http://bit.ly/EL-News
Introduction and rationale
What we do
Typical problems
How many new users complete a funnel?
How many users fail to complete an edit?
Is the conversion rate of a new experimental feature significantly higher than a control?
What is the breakdown by user segment in the adoption of a feature?
open source usable
scalabletransparent
Web analytics
for rapid prototyping and A/B testing
High-quality, granular data to measure how users interact with features and the UI.
client-side events:clicks
impressionsbucketing events
errors
server-side events:transactions
Third-party solutions
Legacy solutions
Legacy solutions
ClickTracking: issues
1. scalability
ClickTracking: issues
2. flexibility
ClickTracking: issues
2. flexibility
ext.articleFeedbackv5@4-option3-cta_learn_more-button_click-overlay
event names
e90RMgpyyOjKArMiyFUadnI1LgqkPXSU|http://en.wikipedia.org/wiki/Main_Page|Foo
additional data
ClickTracking: issues
3. data modeling
No mechanism for data validation
ClickTracking: issues
3. data modeling
Limited support for data documentation and provenance
ClickTracking: issues
4. data-centric collaboration
Product managersData analystsEngineersUX designers
EventLogging: design principles
Support getting desired metrics, not simply all available metrics
Provide data provenance/ enable long-term use of data
Support data-centric collaboration: using a wiki for data modeling
Support versioning/permit data sets to be compared
Support transparency/community scrutiny
Do not capture every action on site, but rich, well-documented data that fit the question
Enforce lightweight client and server-side validation
Creates a contract between data analyst, product manager, engineer around data definitions
Simplify access to data, QA and data analysis
Next• From an idea to a data model: Designing a schema (Maryana)
• Instrumenting code, collecting and validating EventLogging data (Ori)
• Accessing, QA'ing and analyzing EventLogging data (Dario)
